Corneal endothelial cell loss is a well-described complication after anterior segment surgery and has been reported after filtering and non-filtering glaucoma surgeries [18,19,20,21]. Even after standard trabeculectomy with or without mitomycin C, endothelial cell loss of − 9.6 to − 10% have been reported in three different studies at 12 months [18, 22, 23]. The CyPass implant (Alcon Inc., USA) in combination with cataract surgery showed a decrease of − 20.4% in endothelial cell count at 5 years [24]. As a consequence, there was a concern that endothelial cell loss might be a caveat with the use of other implants, e.g., the XEN45 gel stent. Hence, the present study sought to investigate endothelial cell count loss after XEN45 implantation with or without combined cataract surgery during a follow-up of 5 years after surgery.

In both the solo surgery group as well as in the combined surgery group, the observed loss of endothelial cell count was − 1% (95% CI: − 7 to 4%) at 5 years compared to baseline. In general, the observed reduction of endothelial cell count was less or equal to − 4% after 1–5 years of follow-up in the entire study population as well as in the solo and the combined surgery subgroups.

Pairwise comparisons of baseline and postoperative visits revealed a decrease in central endothelial cell counts at 2 and 4 years postoperative overall (− 1%, − 3%) as well as in the combined surgery group (− 3%, − 4%). No statistically significant change was found for 1, 3, and 5 years postoperatively. Our data further suggest a slow, continual decrease of endothelial cell counts over time after combined surgery/cataract surgeries as opposed to a single decrease in endothelial cell count due to surgical trauma without a further decrease of endothelial cell counts over time. We want to point out that solo cataract surgeries can lead to a decrease in central endothelial cell counts (up to − 21% after 11 years) [25, 26]. In the combined surgery cataract group, the median reduction was − 79 (95% CI: − 183 to − 9) at 2 years and − 93 (95% CI: − 220 to − 23) cells at 4 years, corresponding to a mean reduction of − 3% (− 7 to 0) at 2 years and − 4% (− 9 to − 1) at 4 years. Taking into consideration the long follow-up in this study, this seems to be of low clinical relevance, since most XEN45 patients have > 2000 cells preoperatively.

The observed differences between solo and combined surgery groups in the study could be explained by the surgical step of removal of the lens and implanting an IOL, which could put more stress on the endothelial cells in the combined surgery group.

The solo surgery group showed no statistically significant decrease in central endothelial cell counts in all postoperative visits (1–5 years).

Comparison of our data with the XEN45 literature

Olgun et al. reported a loss of − 2.1% (reduction of means from 2156 to 2098, 49 eyes) 3 months after solo surgery implantation [18]. We could not show any significant change in the combined or solo surgery group in our study within the first years postoperatively. The longest retrospective observation period in the literature is provided by Gillman and coworkers who report corneal endothelial cell density over 2 years after combined surgery cataract implantation [27] in which the authors included 17 eyes in their analysis. Gillmann et al. report a decrease of − 14.3% (− 339 cells) at 2 years [27]. There was no statistically significant difference to the solo cataract control group in their study, and whether or not needling revisions were necessary did not influence the decrease of endothelial cells in their study [27]. Our results suggest a smaller decrease of central endothelial cell count over 5 years since we show a decrease of fewer than 130 cells (and less than − 5.4%) at 5 years. The surgical implantation technique was the same in our and their study: especially mitomycin C dosage, amount and route of application, ports setup, and the use of gonioscopy during XEN45 implantation were the same as in their study. However, Gillmann et al. presented no information about the cataract part of the combined surgery. It is relevant to point out that we use a higher frequency of corticosteroid drops in the first week and we prescribed preservative-free eye drops postoperatively. Needling revisions were performed without the use of mitomycin C. Furthermore, the proportion of patients in our study with a loss of more than 20% central endothelial cells postoperatively was lower in comparison to the study by Gillmann et al. They reported a loss of ≥ 20% at 2 years in 3/17 (18%) of the eyes after combined surgery cataract implantation while we detected a decrease of > 20% in 3% of the eyes at 2 years [27].

XEN45 location parameters and association with endothelial cell count change

The distance of the Ahmed tube tip to the cornea has a significant association with endothelial cell loss in the quadrant of implantation [13,14,15]. Tubes that are closer to the cornea seem to lead to increased loss of adjacent endothelial cells (2.5 years after implantation) [12].

Surprisingly, the placement of the XEN45 implant in the anterior chamber angle (e.g., more anterior near the Schwalbe line, more posterior near the scleral spur) and the distance of inner ostium to cornea had no statistically significant influence on the course of central endothelial cell count after XEN45 implantation.

Further to this, after Ahmed valve implantation, the lowest endothelial cell count was recorded at the site of implantation, suggesting that the silicone tube of the Ahmed valve locally reduces endothelial cell counts [28,29,30]. The data of the present study do not support a similar local effect at the entry site of the XEN45 implant. Comparing the endothelial cell count at the implant (measurement superior-nasal) and at the opposite side of the cornea (measurement inferior-temporal) showed a homogenous distribution of endothelial cells over the cornea.

Therefore, the tolerance zone of implantation seems to be large with regard to central endothelial cell loss, possibly giving the surgeon more freedom in terms of positioning the implant at the chamber angle. Despite these findings, we recommend being cautious when choosing the site of implantation, i.e., not implanting too anterior or posterior, as placement could influence other factors (e.g., efficacy and safety: IOP outcome, revision rates, re-surgery rates, needling rates, and the rate of adverse events).

Possible mechanism(s) of loss of central endothelial cell density after glaucoma surgery

The exact mechanism of corneal endothelial damage is unclear [18]. In the case of the silicon tube valve, some authors postulate that jet flow through the silicone tube is caused by the heartbeat, chronic chamber inflammation, mechanical loss by intermittent tube-endothelial cell contact when the patient blinks or rubs the eye, tube-uveal touch, and a foreign body reaction to the silicone might play a role. Further turbulence present at the tip of the implant or microscopic fibrotic tissue between the tube and corneal endothelium may be a possible mechanism for the decrease of endothelial cell count after Ahmed valve implantation [12, 31,32,33]. These reasons might also be causative for a decrease in endothelial cell count after XEN45 implantation. Compared to the low-resistance silicone tubes of plate implants, the XEN45 has a smaller outer diameter, a smaller length in the anterior chamber, and a lower maximal flow rate due to the smaller lumen size, and is made of a less rigid material [18]. This is speculation of course; however, it is not without reason to assume that these factors might account for a beneficial outcome in the endothelial cell count data postoperatively with XEN45 compared to big plate glaucoma devices, which were shown to cause an endothelial cell count loss of − 11.5 to − 18.6% at 24 to 36 months [13,14,15, 34]. Postoperative interventions (e.g., needlings, bleb revisions) do not seem to have any apparent influence on endothelial cell loss after XEN45 implantation.

Whether normal aging, the cataract part of the combined surgery cataract surgery, the XEN45 per se, or a combination of these factors is the reason for the slow decrease in this group, cannot be distinguished with our data. Further investigations and analyses are suggested.

Strengths of the study

To date, a 2-year retrospective follow-up is the longest published dataset. There is no mid-term data on solo surgery procedures published yet. With this study, we investigated the mid-term course of the endothelial cell count on solo surgery procedures.

Limitations of the study

Our study population was strictly Caucasian; therefore, caution is advised in generalization to other ethnicities.

Sample sizes decreased considerably over time. We therefore made an a posteriori power analysis revealing that a decrease in central endothelial cell counts of − 5.4% or more would have been detected by our study design with a power of 87% at the 5-year follow-up. This desirable result achieving high power is mainly due to the fact that endothelial cell counts are highly correlated (r = 0.65 between baseline and 5-year follow-up).

Attrition and loss of follow-up are important issues. However, after checking patients’ records, no evidence for systematic biases (e.g., patients with a lower or higher risk of loss of ECC due to pre-existing conditions or other reasons) was found which may distort study findings. A negative effect of a large loss of follow-up data is that 95% confidence intervals become wider.

No XEN45 location parameter is associated with a higher endothelial cell count loss. Any small clinical effect of XEN45 location parameters on postoperative endothelial cell count course cannot be answered with our data. Therefore, we would suggest follow-up studies with higher sample sizes and further increasing follow-up time.